Epigos Python

Epigos provides an end-to-end platform to annotate data, train computer vision AI models, deploy them seamlessly and host the models via API's.

For more details, visit epigos.ai.

The Epigos Python Package is a python wrapper around the core Epigos AI web application and REST API.

Installation

To install this package, please use Python 3.9 or higher.

To add epigos to your project simply install with pip:

pip install epigos

Or with poetry

poetry add epigos

Getting Started

To make your first API call, you will need to signup at epigos.ai and create an API key for your workspace. Please contact our sales team for a demo.

Initialization:

import epigos

client = epigos.Epigos("api_key")

Project:

Manage project and upload dataset into your project using the Project ID.

Upload an image with annotation

import epigos
from epigos.typings import BoxFormat

client = epigos.Epigos("api_key")

# load project
project = client.project("project_id")

# upload image with Pascal VOC annotation
record = project.upload(
    "path/to/image.jpg",
    annotation_path="path/to/image.xml",
    box_format=BoxFormat.pascal_voc
)
print(record)

# upload image with COCO annotation
record = project.upload(
    "path/to/image.jpg",
    annotation_path="path/to/coco.json",
    box_format=BoxFormat.coco
)
print(record)

# upload image with YOLO annotation
record = project.upload(
    "path/to/image.jpg",
    annotation_path="path/to/image.txt",
    box_format=BoxFormat.yolo
)
print(record)

Upload an entire dataset folder

import epigos

client = epigos.Epigos("api_key")

# load project
project = client.project("project_id")

# upload COCO annotation dataset
records = project.upload_coco_dataset(
    images_directory="path/to/dataset/train/images",
    annotations_path="path/to/dataset/train/coco.json",
)
print(tuple(records))

# upload Pascal VOC annotation dataset
records = project.upload_pascal_voc_dataset(
    images_directory="path/to/dataset/train/images",
    annotations_directory="path/to/dataset/train/labels",
)
print(tuple(records))

# upload YOLO annotation dataset
records = project.upload_yolo_dataset(
    images_directory="path/to/dataset/train/images",
    annotations_directory="path/to/dataset/train/labels",
    data_yaml_path="path/to/dataset/data.yaml",
)
print(tuple(records))

Prediction:

Make predictions with any of the models deployed in your workspace using the Model ID.

Classification

import epigos

client = epigos.Epigos("api_key")

# load classification model
model = client.classification("model_id")

# make predictions
results = model.predict("path/to/your/image.jpg")
print(results.dict())

Object detection

import epigos

client = epigos.Epigos("api_key")

# load object detection model
model = client.object_detection("model_id")

# make predictions
results = model.detect("path/to/your/image.jpg")
print(results.dict())
# visualize detections
results.show()

Contributing

If you want to extend our Python library or if you find a bug, please open a PR!

Also be sure to test your code with the make command at the root level directory.

Run tests:

make test

Commit message guidelines

It’s important to write sensible commit messages to help the team move faster.

Please follow the commit guidelines

Versioning

This project uses Semantic Versioning.

Publishing

This project is published on PyPi

License

This library is released under the MIT License.